Missing symbols in libuuid.a

[joankaradimov]

I'm trying to get a python-pywin32 build running in MinGW. Here is some of my progress. However, I encountered some missing symbols in libuuid.a. They are all present in Uuid.Lib shipped with the Microsoft SDK.

Here's the output from ld:

build/temp.mingw-2.7/com/win32com/src/PythonCOM.o:PythonCOM.cpp:(.rdata$.refptr.IID_IContext[.refptr.IID_IContext]+0x0): undefined reference to `IID_IContext'
build/temp.mingw-2.7/com/win32com/src/PythonCOM.o:PythonCOM.cpp:(.rdata$.refptr.IID_IAccessControl[.refptr.IID_IAccessControl]+0x0): undefined reference to `IID_IAccessControl'
build/temp.mingw-2.7/com/win32com/src/Register.o:Register.cpp:(.rdata+0xe38): undefined reference to `IID_IContext'
build/temp.mingw-2.7/com/win32com/src/Register.o:Register.cpp:(.rdata+0xe60): undefined reference to `IID_IEnumContextProps'
build/temp.mingw-2.7/com/win32com/src/Register.o:Register.cpp:(.rdata+0xeb0): undefined reference to `CLSID_DCOMAccessControl'
build/temp.mingw-2.7/com/win32com/src/extensions/PyIContext.o:PyIContext.cpp:(.rdata$.refptr.IID_IEnumContextProps[.refptr.IID_IEnumContextProps]+0x0): undefined reference to `IID_IEnumContextProps'

[ZachBacon]

chances are a patch will need to be submitted to the mingw-w64 git on sf.net. Shouldn't be hard to make a quick patch for use locally

[joankaradimov]

Could it be a problem with the msys2 distribution? When I grep the mingw-w64 sources I find each of those symbols. On the other hand it's only in headers and some are extern. I don't know if that's enough for the correct static import library to be generated.

On top of that my understanding of COM is limited. Even less so outside MSVC.

[ZachBacon]

It's not an msys2 issue it's a CRT in mingw. Go to file in mingw-w64-crt/libsrc/uuid.c that's where your changes would go for the new uuid's anyways I can probably make the change, but I'm heading to bed myself.

[joankaradimov]

I can probably make the change

It would be great if you can find the time and make it.

How are patches discussed/submitted to mingw-w64? Is it via a mailing list? If I can see an example I could do this myself in the future.

[ZachBacon]

It's submitted via their mailing list.

[ZachBacon]

Got 2 of the UUID's implemented, One of them maybe a little more interesting to implement, but once I get the patches generated I'll send it to the mailing list.

EDIT:

Mind testing these patches against a local mingw-w64 package and see if these work for you? You'll have to build the package from source though.

IID_IContext.patch.txt
IID_IAccessControl.patch.txt

[joankaradimov]

It looks great. There's only one error now...

build/temp.mingw-2.7/com/win32com/src/Register.o:Register.cpp:(.rdata+0xeb0): undefined reference to `CLSID_DCOMAccessControl'

[ZachBacon]

Excellent, I'll work on that today and I'll see about getting that in the mailing list.

[ZachBacon]

A lot of those missing uuid's were actually in objidl.h, so it was a matter of moving them. There was one I had to do a bit of research though and for CSLID_DCOMAccessControl, I'll have to a do a bit of research there too, but it should be easy enough.

[ZachBacon]

This one should fix the CLSID_DCOMAccessControl undefined. If this works, I'll throw it at the mingw-w64 devs in their mailing list, I honestly don't know if they'll accept it, and I'm not sure @Alexpux would accept the patches in the MinGW CRT package either
CLSID_DCOMAccessControl.patch.txt

[Alexpux]

Create pull request for mingw-w64 headers/crt packages and I will review

[ZachBacon]

Alright, as soon as @joankaradimov tests to see if it fixes his pywin32 package. Then I'll be submitting these patches for mingw-64's mailing list and then I'll make the PR for the mingw crt msys package.

[joankaradimov]

If this works, I'll throw it at the mingw-w64 devs in their mailing list

The last patch works too. And it got me through that step of the compilation.

Alright, as soon as @joankaradimov tests to see if it fixes his pywin32 package

Well, it doesn't fix the package, but it does fix the compilation of one of the python modules in it :)
I'm already working on the rest.

Create pull request for mingw-w64 headers/crt packages and I will review

@Alexpux: I don't know if that's necessary. I'm not in a hurry. There are quite a few other issues to fix first in pywin32. We can wait and see if this will be merged in master. Right now I have a package that works locally and that's good enough for now.

[ZachBacon]

@joankaradimov I'll be doing the PR for the packages, because it'll be nice to have that all situated as long as it works. Helping you is actually helping me contribute back to the mingw-w64 project as a whole. Besides fixing those uuid's I'm working on adding some missing includes for OBS Studio that mingw-w64 is missing like wscapi and iwscapi as well as some mmdevapi speaker configs. So like I said, l'll deal with this, as it's a learning experience for me :)

[joankaradimov]

@ZachBacon: What you are saying makes sense.

While getting more modules to compile, I encountered another missing symbol - IID_IHTMLOMWindowServices

[joankaradimov]

... and some more: IID_IEmptyVolumeCacheCallBack, IID_IEmptyVolumeCache, IID_IEmptyVolumeCache2, FMTID_ImageSummaryInformation, FMTID_MediaFileSummaryInformation, FMTID_AudioSummaryInformation, CLSID_CTask, CLSID_CTaskScheduler, IID_IProvideTaskPage, IID_IScheduledWorkItem, IID_ITask, IID_ITaskScheduler, IID_ITaskTrigger.

There are a lot of those, it appears :/

@ZachBacon Maybe I should take over. I'll understand if you don't have the time to deal with all of these.

[ZachBacon]

@joankaradimov You underestimate the amount of free time I have now adays. The sad benefits of leaving ones job and in the middle of looking for a new one. Nah, I got this. Though if you do want to help, I've forked the mingw-w64 repo at https://github.com/ZachBacon/mingw-w64 I'm occasionally syncing with upstream so that I can make the proper changes and send them signed off patches. I can give you commit access there, that way we can both work on it.

[joankaradimov]

I can give you commit access there, that way we can both work on it.

No need. It's better to open PRs in your repo. This way you can review the changes.

Maybe I can start backwards alphabetically and do IID_ITask* stuff first.

[Alexpux]

Some GUIDS:

3050F5FC-98B5-11CF-BB82-00AA00BDCE0B IHTMLOMWindowServices
6E793361-73C6-11D0-8469-00AA00442901 IEmptyVolumeCacheCallBack
02B7E3BA-4DB3-11D2-B2D9-00C04F8EEC8C IEmptyVolumeCache2
8FCE5227-04DA-11D1-A004-00805F8ABE06 IEmptyVolumeCache
4086658A-CBBB-11CF-B604-00C04FD8D565 IProvideTaskPage
A6B952F0-A4B1-11D0-997D-00AA006887EC IScheduledWorkItem
148BD524-A2AB-11CE-B11F-00AA00530503 ITask
148BD527-A2AB-11CE-B11F-00AA00530503 ITaskScheduler
148BD52B-A2AB-11CE-B11F-00AA00530503 ITaskTrigger

[Alexpux]

in other form:

DEFINE_GUID(IID_IHTMLOMWindowServices, 0x3050F5FC, 0x98B5, 0x11CF, 0xBB, 0x82, 0, 0xAA, 0, 0xBD, 0xCE, 0xB);
DEFINE_GUID(IID_IEmptyVolumeCacheCallBack, 0x6E793361, 0x73C6, 0x11D0, 0x84, 0x69, 0, 0xAA, 0, 0x44, 0x29, 0x1);
DEFINE_GUID(IID_IEmptyVolumeCache2, 0x02B7E3BA, 0x4DB3, 0x11D2, 0xB2, 0xD9, 0, 0xC0, 0x4F, 0x8E, 0xEC, 0x8C);
DEFINE_GUID(IID_IEmptyVolumeCache, 0x8FCE5227, 0x04DA, 0x11D1, 0xA0, 0x4, 0, 0x80, 0x5F, 0x8A, 0xBE, 0x6);
DEFINE_GUID(IID_IProvideTaskPage, 0x4086658A, 0xCBBB, 0x11CF, 0xB6, 0x4, 0, 0xC0, 0x4F, 0xD8, 0xD5, 0x65);
DEFINE_GUID(IID_IScheduledWorkItem, 0xA6B952F0, 0xA4B1, 0x11D0, 0x99, 0x7D, 0, 0xAA, 0, 0x68, 0x87, 0xEC);
DEFINE_GUID(IID_ITask, 0x148BD524, 0xA2AB, 0x11CE, 0xB1, 0x1F, 0, 0xAA, 0, 0x53, 0x5, 0x3);
DEFINE_GUID(IID_ITaskScheduler, 0x148BD527, 0xA2AB, 0x11CE, 0xB1, 0x1F, 0, 0xAA, 0, 0x53, 0x5, 0x3);
DEFINE_GUID(IID_ITaskTrigger, 0x148BD52B, 0xA2AB, 0x11CE, 0xB1, 0x1F, 0, 0xAA, 0, 0x53, 0x5, 0x3);

[smalltalkman]

Not tested, but may be helpful:

DEFINE_GUID(CLSID_CTask, 0x148BD520, 0xA2AB, 0x11CE, 0xB1, 0x1F, 0x00, 0xAA, 0x00, 0x53, 0x05, 0x03);
DEFINE_GUID(CLSID_CTaskScheduler, 0x148BD52A, 0xA2AB, 0x11CE, 0xB1, 0x1F, 0x00, 0xAA, 0x00, 0x53, 0x05, 0x03);

from winsdk-10

// FMTID_ImageSummaryInformation
#define PSGUID_IMAGESUMMARYINFORMATION {0x6444048fL, 0x4c8b, 0x11d1, 0x8b, 0x70, 0x8, 0x00, 0x36, 0xb1, 0x1a, 0x03}
// FMTID_MediaFileSummaryInformation, see propidl.h PID_ values for this
#define PSGUID_MEDIAFILESUMMARYINFORMATION {0x64440492L, 0x4c8b, 0x11d1, 0x8b, 0x70, 0x08, 0x00, 0x36, 0xb1, 0x1a, 0x03}

from winsdk-10

FMTID_AudioSummaryInformation     : TGuid = '{64440490-4C8B-11D1-8B70-080036B11A03}';

from here

[ZachBacon]

excellent, I'll get those in later tonight or tomorrow, if things test out well, I'll send in the patches as a unified patch to the mingw-w64 project as well.

What I'd like to do eventually, any problematic packages that won't compile due to something missing in mingw-w64 but compiles with msvc, I'd like to try and make patches to the mingw-w64 sdk, help expand the compatibility and expand what we can carry in the repository potentially.

[ZachBacon]

Got the UUID's in, Gonna work on the FMTID_AudioSummaryInformation in next
https://github.com/ZachBacon/mingw-w64/commit/d9fa05738de7de26147e460c29ac80b088422799.patch

[ZachBacon]

Apologies for delays, other things kinda took my attention. I'll submit a PR soon regarding the missing uuid's

[joankaradimov]

Apologies for delays, other things kinda took my attention.

No problem. Thank you for getting involved in this!

I'll submit a PR soon regarding the missing uuid's

Could you review my PR in your repo and potentially squash it into you changes?

[ZachBacon]

yeah I could do that.

[ZachBacon]

And patches sent to the mingw-w64-public mailing lists

[ZachBacon]

Sometime this week, I'm going to attempt to make a PR for the patches we made for the headers/crt packages, due to the fact they are now pushed to the mailing list for mingw-w64, and this should now allow the pywin32 package to be compiled for msys2

[ZachBacon]

And got confirmation, the patches got applied upstream.

[joankaradimov]

They have also landed in MSYS2 with commit 8146fdb. This is fixed. I'm closing the issue.

Thanks, @ZachBacon, and thanks everyone!